The Keck/DEIMOS Stellar Archive: I. Uniform Velocities and Metallicities for 78 Milky Way Dwarf Galaxies and Globular Clusters

TL;DR

This work provides a homogeneous catalog of radial velocities and Ca II triplet-based metallicities for 22,339 stars across 78 Milky Way dwarf galaxies and globular clusters, all observed with Keck/DEIMOS 1200G. A novel forward-modeling velocity method (dmost) coupled with per-slit telluric templates and a data-driven LSF yields high-precision velocities with a validated error floor of 1.1 km s$^{-1}$, enabling improved dynamical mass estimates. The authors implement a rigorous EW-based metallicity framework, robust membership probabilities combining photometry, kinematics, and Gaia data, and provide extensive merged catalogs plus extragalactic redshifts, all accessible via KOA. Validation against Gaia, APOGEE, DESI, and MMT/Hectochelle confirms small velocity zeropoints and realistic uncertainties, while the approach reduces systematic biases that plagued earlier DEIMOS analyses. This dataset underpins a companion study (Paper II) that derives dynamical masses, mean metallicities, and metallicity spreads for the MW satellite population, advancing tests of dark matter and galaxy formation at the low-mass end.

Abstract

We present a homogeneous spectroscopic dataset of 22,339 stars in 78 Milky Way dwarf galaxy satellites and globular clusters. All data were taken with the Keck II telescope and Deep Extragalactic Imaging Multiobject Spectrograph (DEIMOS) spectrograph using the 1200G grating (spectral resolution R~6000). Based on a uniform data reduction of 411 DEIMOS masks, we present a catalog of individual stellar radial velocities, equivalent width-based [Fe/H] metallicities, and membership estimates. The Milky Way satellites range from M_V = 2 to -14 (M* = 10^1.5 to 10^7.5 Msun); the majority of individual stars presented in these systems have magnitudes 17 > r > 22. The data were reduced to 1D spectra using PypeIt, which provides near Poisson statistics-level sky subtraction. Radial velocities were determined via dmost, a forward modeling method first presented here, which combines both synthetic telluric and stellar templates to determine stellar radial velocities. We assess the accuracy and precision our method via comparison to thousands of repeat measurements and literature values. We determine a velocity error floor of 1.1 km/s and a CaII triplet-based metallicity error floor of 0.1 dex. We calculate internal velocity dispersions and compare to literature values, demonstrating 20-50% improved precision over the literature in most cases. In a companion paper, we use our homogeneous catalogs to explore properties of these Milky Way satellites, including previously unpublished measurements in several systems including Bootes II and Draco II. We provide full access to the data catalogs to enable further studies.

Figures (18)

• Figure 1: Spatial distribution (RA/Dec) of Milky Way (MW) dwarf satellite galaxies (blue), globular and faint star clusters (green), and M31 dwarf galaxy satellites (red) observed with the Keck/DEIMOS instrument and 1200G grating. The circle size represents the number of 1200G Keck/DEIMOS archival pointings (unique multiobject masks) for each object, ranging between 1 and 26. The median number of DEIMOS masks per object is three. The black solid line is the Galactic plane; the gray shaded area is the region of sky not visible from the Keck telescope latitude. Spectra extracted from all MW systems are presented in this paper. M 31 will be presented in a future contribution.
• Figure 2: An example DEIMOS 1200G science spectrum (black) for a high S/N star in the Draco dSph galaxy ($r=16.5$, $t_{\rm exp} = 1200$ s, S/N$\sim$80). We plot the associated sky emission spectrum (blue), scaled down for plotting purposes. Orange squares indicate relatively isolated sky emission lines used in the flexure fitting (§ \ref{['ssec:flexure']}). The wavelength region used to fit telluric absorption quantities and the primary telluric absorption species are indicated above each spectral region. The red line is the best-fitting dmost forward model spectrum created by combining a synthetic stellar and synthetic telluric absorption spectrum.
• Figure 3: To correct for wavelength flexure between the afternoon arc calibrations and evening science exposures, we fit Gaussian line profiles to a set of isolated sky emission lines across all slits. Left: The difference between the predicted and observe sky emission line center ($\Delta \lambda_f$) versus wavelength before flexure correction for a two slits on the edge ( top) and center ( bottom) of the mask. Non-zero wavelength offsets of skylines in the science exposure are corrected during the flexure procedure. Right: Gaussian line width as a function of wavelength. Although the LSF depends slightly on wavelength, we opt to use the average value as the LSF for each slit.
• Figure 4: The flexure fit values for all slits in a single DEIMOS exposure based on sky emission lines. We plot the slope ($m_f$, left), intercept ($b_f$, middle) and average line width (LSF, right) as a function of mask position. The pattern in the left two panels is due to flexure between the afternoon wavelength calibration and science exposure, while the rightmost panel is due to variation in image quality across the spectrograph.
• Figure 5: To determine the best atmospheric telluric absorption parameters for a single DEIMOS exposure, we determine values in all slits by searching over a grid of synthetic telluric spectra. We plot the inferred humidity (H$_2$O) and the oxygen mixing ratio (O$_2$) as a function of the median S/N in each slit. The telluric values for each mask are found by the weighted average of the slit values (horizontal blue line in each panel).